Progressive Backmapping of Highly Coarse-Grained Protein Models

Area of Science:

• Computational Biology
• Molecular Dynamics
• Biophysics

Background:

• Reconstructing all-atom (AA) structures from coarse-grained (CG) models is crucial for multiscale molecular dynamics (MD) simulations of large biomolecular assemblies.
• Current methods struggle with mesoscale systems beyond conventional MD reach.

Purpose of the Study:

• To develop a progressive backmapping framework for accurate AA structure reconstruction from CG models.
• To enable detailed analysis of complex mesoscale biomolecular systems.

Main Methods:

• Introduced a stepwise, progressive backmapping approach building upon neural-network-based ProNet Backmapping.
• Implemented reconstruction across neighboring resolutions (e.g., 3-residue-per-site HCG to AA).
• Applied hierarchical reconstruction to complex protein assemblies and viral particles.

Main Results:

• Achieved high accuracy in reconstructing AA models for diverse proteins and flexible linkers.
• Successfully reconstructed large, complex protein assemblies, including multiple virus-like particles.
• Demonstrated hierarchical backmapping of entire viral assemblies from HCG to AA resolution across three levels.

Conclusions:

• The progressive backmapping framework offers a scalable solution for incorporating atomistic detail into mesoscale simulations.
• This method advances the study of complex biological systems in chemistry and biology.